Putting sample inputs with their outputs into a prompt to demonstrate the pattern and format wanted, without altering model weights; zero-shot prompting gives no samples.
Also called few-shot prompting.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains Few-shot learning in context, with comparison tables and the common traps.
Terms in this definition
- Pattern
A reusable answer to a recurring problem that places building blocks in context, explaining when, how and why to apply them and what compromises are involved.
- FORMAT
A DAX function that turns a value into text according to a format string, for instance "MMMM" to show a month's name. Since the output is text, numeric operations can't use it, and dynamic format strings were introduced to get around that.
- Model parameters
The weights a model learns in training, which together hold what it knows. How many there are is what distinguishes a small language model from a large one.
- Zero-shot prompting
Giving a model only the instruction without any examples, for instance asking a chat model to sort text into categories named in the prompt.